User not logged in - login - register
Home Calendar Books School Tool Photo Gallery Message Boards Users Statistics Advertise Site Info
go to bottom | |
 Message Boards » » AI/Machine Learning/Neural Net APIs/libs/tools etc Page [1]  
Data_
Veteran
311 Posts
user info
edit post

Post here (preferably projects that are still being maintained):

The OpenCog Framework - http://wiki.opencog.org
Emergent Neural Network Simulation System -http://grey.colorado.edu/emergent/index.php/Main_Page
Neuroph Java Neural Network Framework - http://neuroph.sourceforge.net
Encog Machine Learning Framework - http://www.heatonresearch.com/encog

These are just a few of the best I've found. If you know of any good ones, post em!

2/27/2013 3:14:29 PM

neolithic
All American
706 Posts
user info
edit post

It depends up what you're trying to do I guess. I use a lot of machine learning techniques, but they aren't for neural simulation, so I use R for most of my work.

http://cran.r-project.org/web/views/MachineLearning.html

My favorite packages include gradient boosting machines (gbm package), random forests (randomForest), and multivariate adaptive regression splines (earth). I'm not a huge fan of the neural network packages in R. I use a lot of these as base learners that I then combine or "ensemble" them to form higher level, meta-learners.

Deep learning has received quite a bit of attention this year. Deep learning refers to training multi-layer neural networks (i.e. "deep" nets) that are capable of learning complicated structures. There is a nice toolbox for matlab and a site dedicated to deep learning techniques.

http://deeplearning.net/
http://deeplearning.net/software_links/

They have come to dominate speech recognition, at least that's my understanding. The press has picked up on some of google's deep learning experiments:

http://deeplearning.net/2012/12/13/googles-large-scale-deep-learning-experiments/

There is an excellent free online class (not a MOOC) going on at NYU right now, but it's not based on neural nets.

http://cilvr.cs.nyu.edu/doku.php?id=courses:bigdata:start

There was a really good class on neural networks on coursera taught by Hinton, who is one of the major pioneers in the field, but I don't know if they plan to teach it again.

If you want to get your hands dirty, pick your favorite method and try one of the contests on Kaggle:

http://www.kaggle.com/

I know that python is getting popular too. Numpy and Scipy get used a lot and implement a variety of techniques, but I've come to dislike python, so I rarely use it.

[Edited on February 27, 2013 at 4:07 PM. Reason : ]

2/27/2013 4:04:08 PM

Data_
Veteran
311 Posts
user info
edit post

Oh damn I had forgotten about deep learning. I remember hearing about it and thinking "oh well that's not new" and for that reason I guess I forgot about it never did much research on it. I won't make that mistake again.

2/27/2013 4:14:17 PM

neolithic
All American
706 Posts
user info
edit post

Hinton gave a good talk about deep learning at NIPS this year:

http://videolectures.net/nips2012_hinton_networks/

It's about a technique to improve them, but should get you up to speed on deep nets as well. Might as well check out the other NIPS talks while you're at it:

http://videolectures.net/nips2012_laketahoe/

[Edited on February 27, 2013 at 4:21 PM. Reason : ]

Oh yeah, one of the best technical books out there is available as a PDF for free as well.

http://www-stat.stanford.edu/~tibs/ElemStatLearn/
http://www-stat.stanford.edu/~tibs/ElemStatLearn/printings/ESLII_print10.pdf



[Edited on February 27, 2013 at 4:47 PM. Reason : ]

2/27/2013 4:20:51 PM

Data_
Veteran
311 Posts
user info
edit post

Thanks. VideoLectures has filled many idle hours in the past, but I had not actually found anything particularly interesting recently. So this is good.

2/27/2013 4:46:29 PM

neolithic
All American
706 Posts
user info
edit post

Don't know how I forgot about NELL:

http://rtw.ml.cmu.edu/rtw/

2/27/2013 8:12:58 PM

Data_
Veteran
311 Posts
user info
edit post

bump! I was hoping more people had some stuff. Oh well

3/3/2013 11:13:56 PM

Stein
All American
19842 Posts
user info
edit post

http://conductrics.com/data-science-resources/
http://conductrics.com/data-science-resources-2

3/4/2013 9:09:58 AM

BigMan157
Suspended
103036 Posts
user info
edit post

there's some library i used for making neural networks in matlab once upon a time, but i've pretty much forgotten it at this point

fuck, let me look at some old code and see if i can find a function call i can google

[Edited on March 4, 2013 at 9:39 AM. Reason : found it http://www.mathworks.com/products/neural-network/]

3/4/2013 9:37:30 AM

neolithic
All American
706 Posts
user info
edit post

^To build on that, I'm using a deep learning toolbox currently.
https://github.com/rasmusbergpalm/DeepLearnToolbox

Vowpal Wabbit (It's pronounced in the way Elmer Fudd would pronounce Vorpal Rabbit) is good for big data sets: https://github.com/JohnLangford/vowpal_wabbit/wiki

Weka is a popular Java machine learning platform: http://www.cs.waikato.ac.nz/ml/weka/

Everything I've posted in this thread so far would largely fall under probabilistic methods, meaning they attempt to find assocations between inputs and outputs. There is a large literature on finding causal relationships. Judea Pearl is the big name in this field. Here is his blog:

http://www.mii.ucla.edu/causality/

and his seminal book on the topic:

http://www.amazon.com/gp/product/052189560X/ref=pd_lpo_k2_dp_sr_1?pf_rd_p=486539851&pf_rd_s=lpo-top-stripe-1&pf_rd_t=201&pf_rd_i=0521773628&pf_rd_m=ATVPDKIKX0DER&pf_rd_r=0B0GY7PT27CB569XQPXW

Here are some blogs I frequently read, that are at least related to ML:
http://hunch.net
http://normaldeviate.wordpress.com
http://nuit-blanche.blogspot.it
http://rjlipton.wordpress.com
http://www.scottaaronson.com/blog/

3/4/2013 10:50:30 AM

neolithic
All American
706 Posts
user info
edit post

If you're looking for some data to try these methods on, check out:

http://archive.ics.uci.edu/ml/
http://www.sigkdd.org/kddcup/index.php
http://datamob.org
http://www.fedstats.gov
http://www.census.gov/main/www/access.html

If you really want to get some experience using these methods, I'll recommend again you sign up at Kaggle and enter some competitions.

Do you have any specific interests or projects you are currently or would like to work on?

3/4/2013 11:02:02 AM

DeltaBeta
All American
9414 Posts
user info
edit post

InsaneMan?

3/4/2013 11:52:43 AM

neolithic
All American
706 Posts
user info
edit post

Julia is supposed to be the next big thing in scientific computing, but right now it's still pretty immature:

http://julialang.org/

3/4/2013 2:50:47 PM

neolithic
All American
706 Posts
user info
edit post

Not a library, but related:

http://www.nytimes.com/2013/02/18/science/project-seeks-to-build-map-of-human-brain.html?pagewanted=1&_r=0

3/5/2013 10:10:24 AM

shoot
All American
6423 Posts
user info
edit post

The time for mobility is ending, the time for AI is coming!

10/9/2018 3:08:51 PM

moron
All American
31777 Posts
user info
edit post

neolithic was ahead of his time, a lot of those resources still hold up today.

10/9/2018 9:50:22 PM

 Message Boards » Tech Talk » AI/Machine Learning/Neural Net APIs/libs/tools etc Page [1]  
go to top | |
Admin Options : move topic | lock topic

© 2019 by The Wolf Web - All Rights Reserved.
The material located at this site is not endorsed, sponsored or provided by or on behalf of North Carolina State University.
Powered by CrazyWeb v2.37 - our disclaimer.